Abstract: A high-quality, long-series, real-time updatable snow depth dataset of in-situ observations collected by National Meteorological Information Centre is developed. It is crucial for evaluating snow depth model and satellite remote sensing products. A comprehensive quality control procedure including the check of metadata, limit values, temporal consistency, temperature-snow depth consistency, precipitation-weather phenomenon-temperature-snow depth consistency, and spatial consistency has been applied to snow depth in-situ observations to identify erroneous data. About 0.2% erroneous data are identified, especially the false “0” value and the erroneous extremes at 148 stations, and the quality of the dataset is ensured. The dataset comprises of snow depth observations from about 2400 sites spanning from 1951 to 2023, with data completeness ranging from 80%—90% at each station. Based on this dataset, the present study investigates spatial distribution of snow in different seasons, the three major snowpack zones, extreme values, and climatic trend of mean daily snow depth and the number of snow-cover days. The findings highlight the largest snow depth and the number of snow-cover days in specific regions like Northeast China, the eastern Inner Mongolia, northern Xinjiang, and the Qingzang Plateau (15—20 cm and over 80 days in winter). Spatial variability is also observed on the Qingzang Plateau. Significant increasing trends are observed in Northeast China, the eastern Inner Mongolia, and northern Xinjiang, while decreasing trends are observed in the North China Plain and the Qingzang Plateau.